Health & Wellness
Repetitive intake and triage work, under data handling constraints that have to be settled first.
Read the second half of this page before the first.
Where the cost is
Health workloads are unusually repetitive: intake, triage, structured extraction from documents, and summarisation, all following a small number of templates. That is exactly the shape Cortex learns from, and exactly the shape where a smaller model, given a tighter prompt, holds its accuracy.
Where the constraint is
Learning a workload means storing its content. Once a workload recurs, Operant stores the text of its messages, tool inputs and tool outputs by default. If that content is health information, that is a decision to make deliberately and in writing — not one to discover in section 5.3 of the Privacy Policy after the fact.
Three things you should know before routing anything sensitive:
- Escalated capture is on by default, and you can ask us to turn it off for your account.
- Our redaction pass is pattern matching, not classification. It will not catch a condition described in prose. Do not rely on it as de-identification.
- We hold no health-sector certification. Security lists what we have and have not built.
Operant can be run against this workload with capture disabled — you lose skill learning and keep routing, compression and the Ledger. That is often the right first step here.
Where to start
A waste audit answers this with your numbers rather than ours. Nothing to install, and you keep the findings either way.
This page describes where cost concentrates in this kind of workload. It is not a claim about customers in this sector — Operant is pre-launch and has none to name.